cambridge.org/psm
Original Article
Cite this article:Meyer N, Joyce DW, Karr C, de Vos M, Dijk D-J, Jacobson NC, MacCabe JH (2021). The temporal dynamics of sleep disturbance and psychopathology in psychosis: a digital sampling study. Psychological Medicine 1–10. https://doi.org/ 10.1017/S0033291720004857
Received: 2 June 2020 Revised: 28 October 2020 Accepted: 26 November 2020
Key words:
Digital health; experience sampling; insomnia; mHealth; relapse prediction; schizophrenia
Author for correspondence: Nicholas Meyer,
E-mail:nicholas.meyer@kcl.ac.uk
© The Author(s), 2021. Published by Cambridge University Press. This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
and psychopathology in psychosis: a digital
sampling study
Nicholas Meyer1,2 , Dan W. Joyce3, Chris Karr4, Maarten de Vos5,6, Derk-Jan Dijk7,8, Nicholas C. Jacobson9and James H. MacCabe1,2
1Department of Psychosis Studies, Institute of Psychology, Psychiatry and Neuroscience, King’s College London,
London, UK;2National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London, London, UK;3
Department of Psychiatry, National Institute of Health Research, Oxford Health Biomedical Research Centre, Warneford Hospital, University of Oxford, Oxford, UK;4Audacious Technologies, Chicago, IL, USA;5Institute of Biomedical Engineering, University of Oxford, Oxford, UK;6ESAT, Department of Engineering & Department of Development and Regeneration, KU Leuven, Leuven, Belgium;7Sleep Research Centre, University of Surrey, Surrey, UK;8UK Dementia Research Institute, London, UK and9Center for Technology and Behavioral Health, Geisel School of Medicine, Dartmouth College, Lebanon, NH, USA
Abstract
Background.Sleep disruption is a common precursor to deterioration and relapse in people living with psychotic disorders. Understanding the temporal relationship between sleep and psychopathology is important for identifying and developing interventions which target key variables that contribute to relapse.
Methods.We used a purpose-built digital platform to sample self-reported sleep and psycho-pathology variables over 1 year, in 36 individuals with schizophrenia. Once-daily measures of sleep duration and sleep quality, and fluctuations in psychopathology (positive and negative affect, cognition and psychotic symptoms) were captured. We examined the temporal rela-tionship between these variables using the Differential Time-Varying Effect (DTVEM) hybrid exploratory-confirmatory model.
Results.Poorer sleep quality and shorter sleep duration maximally predicted deterioration in psychosis symptoms over the subsequent 1–8 and 1–12 days, respectively. These relationships were also mediated by negative affect and cognitive symptoms. Psychopathology variables also predicted sleep quality, but not sleep duration, and the effect sizes were smaller and of shorter lag duration.
Conclusions.Reduced sleep duration and poorer sleep quality anticipate the exacerbation of psychotic symptoms by approximately 1–2 weeks, and negative affect and cognitive symptoms mediate this relationship. We also observed a reciprocal relationship that was of shorter dur-ation and smaller magnitude. Sleep disturbance may play a causal role in symptom exacerba-tion and relapse, and represents an important and tractable target for intervenexacerba-tion. It warrants greater attention as an early warning sign of deterioration, and low-burden, user-friendly digital tools may play a role in its early detection.
Introduction
Sleep disturbance is a common feature of the relapse prodrome of schizophrenia, reported in between 43% and 70% of patients (Birchwood et al.,1989; Heinrichs & Carpenter,1985; Herz & Melville,1980; Jorgensen,1998), and appearing in the weeks prior to the onset of significant deterioration in mental state. These observations are clinically relevant, as they raise the pos-sibility that dysregulated sleep is an important early warning sign of deterioration. If this were to be the case, sleep dysfunction may also play a causal role in psychotic deterioration and relapse, and present a tractable target for preventative interventions which avert progression to relapse. However, no studies have specifically examined the time-course of sleep disturbance in relation to the deterioration of psychosis.
Recent clinical studies have begun to examine the hypothesised association between sleep and psychopathology. In a 7-day experience sampling method (ESM) study of self-reported and actigraphic sleep variables on next-day symptoms in schizophrenia, Mulligan, Haddock, Emsley, Neil, and Kyle (2016) found that poorer sleep predicted greater next-day negative affect and psychosis symptoms, with negative affect mediating a proportion of this relation-ship. Similarly, in a 6-day ESM study in individuals on the psychosis-spectrum, Kasanova, Hajdúk, Thewissen, and Myin-Germeys (2020) reported a significant effect of poor sleep qual-ity on paranoia the next morning, with negative affect fully mediating the relationship. Here, no evidence of an inverse relationship– greater paranoia on evening sleep quality – was found.
A third study by Reeve, Nickless, Sheaves, and Freeman (2018) assessed subjective insomnia, psychosis and negative affect symp-toms at monthly intervals over 3 months, in 29 patients with non-affective psychosis. Insomnia predicted later psychosis symptoms, and this relationship was again mediated by negative affect. This study also found a reciprocal effect between psychotic symptoms and insomnia, implying a bidirectional relationship between these variables.
Although canonical ESM offers high temporal resolution by intensively sampling mental state at multiple time-points through-out the day, the approach has some limitations. It is demanding of the participant, and lends itself to short periods of observation (Palmier-Claus et al., 2011). Analytic approaches have typically used multilevel models to examine short-term associations between variables from one day to the next, and are not optimised for exploring lag structures over longer timescales. Studies are required that are of sufficient duration to capture longitudinal fluctuations in sleep and psychopathology, while of adequate temporal resolution to sample variation over a range of timescales. Methods of analysis that allow the temporal structure between longitudinal variables to be examined are also necessary.
Designed collaboratively with service users, the Sleepsight plat-form (Meyer et al.,2018) capitalises on the ubiquity and usability of smartphone technologies to allow long-term sampling of self-reported sleep (sleep quality and duration) and psychopathology variables (positive affect, negative affect, cognition and psychosis symptom) in individuals living with psychosis, with minimal bur-den to the user. We captured these variables once a day over the course of a year, in 36 individuals with schizophrenia, collecting over 12 000 data-points.
We employed an analytic tool developed specifically for investi-gating the temporal structure of intensive longitudinal data, the Differential Time-Varying Effect Model (DTVEM; Jacobson, Chow, & Newman, 2019). DTVEM examines the lag structure between one or more time-series, using a two-stage, exploratory– confirmatory model. This approach examines the lags at which one construct maximally predicts itself, or another construct, at a later time, and here allows the strength and timing of association between sleep and psychopathology to be studied (Fig. 1).
We predicted that shorter sleep duration and poorer sleep quality would be associated with greater intensity of psychotic, negative affect and cognitive symptoms. In the context of findings supporting a bidirectional relationship between sleep disturbances and psychopathology from the mood disorder literature (Talbot et al.,2012), we also expected to observe a reciprocal relationship between psychopathology and sleep, but predicted that this asso-ciation would be of shorter duration and lower magnitude. In line with previous research, we also hypothesised that the relationship between sleep and psychosis symptoms would be mediated by negative affect but also cognitive symptoms, as evidenced by a weaker and less durable direct effect of sleep on psychosis when these variables are included as mediators.
Methods
Sleepsight app and platform
Developed specifically to optimise acceptability and feasibility in people with psychosis, the Sleepsight system consisted of an Android smartphone application and secure back-end infrastruc-ture, for the real-time collection of self-reported sleep and psycho-pathology variables. Further details of the development and
testing of the system (Biagianti, Hidalgo-Mazzei, & Meyer, 2017; Meyer et al., 2018) and source code for the application
have been published elsewhere (Karr, 2017). Although
Sleepsight also gathered continuous objective rest-activity vari-ables from smartphone sensors and wearable devices, in the pre-sent paper, we focus uniquely on the self-reported, subjective variables.
A fixed sampling schedule prompted users to complete a brief once-daily sleep and symptom diary, between 11:00 and 14:00. This window was chosen to be early enough to minimise forget-ting the previous night’s sleep times, while also avoiding diurnal extremes of mood (Peeters, Berkhof, Delespaul, Rottenberg, & Nicolson,2006). If not completed after the first notification, par-ticipants received two further prompts at 30 min intervals. The sleep diary (Fig. 2b) consisted of four items: time into bed, time out of bed, sleep duration and sleep quality, rated from 1 (‘very poor’) to 5 (‘very good’). All times were entered using a 12 h clock, with AM and PM specifiers, to the nearest 15 min. The symptom diary (Fig. 2c, d) consisted of items validated and widely used in previous ESM studies of psychosis (e.g. Myin-Germeys, Van Os, Schwartz, Stone, & Delespaul, 2001; Palmier-Claus et al. 2012). Rated on a seven-point Likert scale anchored at ‘not at all’ to ‘very much’, they consisted of seven mood items (‘over the past 24 h, I have been feeling: cheerful, anxious, relaxed, irritable, sad, in control, stressed’) and seven items relating to psychosis and cognition (‘over the past 24 h, I have been experi-encing/feeling: suspicious, trouble concentrating, preoccupied by thoughts, others dislike me, confused, others influence my thoughts, unusual sights and sounds’). Positive and negative mood items were counterbalanced, to reduce skew and minimise stereotyped responses. The diary took approximately 1–3 min to complete, and all sections had to be completed in order for the app to allow submission. Participants received a motivational message on submitting a diary, but did not receive feedback based on their entries.
Participants and study procedures
Adults aged 18–65, living in the community with a diagnosis of schizophrenia or schizoaffective disorder, and under the care of secondary care mental health services were invited to take part. Participants were required to be able to provide informed consent and interact comfortably with the smartphone interface, but were otherwise not excluded on the severity of their psychosis symp-toms or medication status.
On study entry, the Positive and Negative Syndrome Scale (PANSS), Pittsburgh Sleep Quality Index (PSQI; Buysse,
Reynolds, Monk, Berman, & Kupfer, 1989) and Insomnia
Severity Index (ISI; Bastien, Vallières, & Morin,2001) were admi-nistered. Participants without their own device were provided with an Android smartphone, which they were asked to use as their sole device throughout the study period. For eligible participants already in possession of an Android device, the Sleepsight app was installed on their own device. Therefore, the risk of non-response by having multiple devices was minimised, by ensuring all participants only made use of one device. All participants received equal compensation for their time, regardless of device ownership status, and received a face-to-face training session on enrolment.
Participants were asked to respond to the once-daily smartphone notification by completing the smartphone diary, and use their mobile devices as normal, for 12 months. Participants received a
weekly motivational SMS message, and were contacted if they were having technical difficulties, or if the engagement was low. Otherwise, no change was made to the delivery of psychiatric care. Screening for eligibility and all clinical assessments were under-taken by an experienced psychiatrist (NM), and the study received ethical approval from the local Research Ethics Committee.
Data analysis
Data were organised with the unit of time as one day. Four sum-mary psychopathology variables were computed from diary items: positive affect (mean of‘cheerful, ‘relaxed’ and ‘in control’); nega-tive affect (mean of‘anxious’, ‘irritable’, ‘sad’, ‘stressed’), cognitive
symptoms (mean of ‘trouble concentrating’, ‘confused’) and
psychosis symptoms (mean of ‘suspicious’, ‘preoccupied by
thoughts’, ‘others dislike me’, ‘others influence my thoughts’ and ‘unusual sights and sounds’). In order to examine associa-tions between sleep and individual psychotic symptoms in more detail, two additional symptoms were included in a secondary analysis: hallucinations (i.e. the ‘unusual sights and sounds’ item) and paranoia (mean of‘suspicious’ and ‘others dislike me’ items). Duplicate entries within a 24 h period were removed, and summary variables visualised (Fig. 3). Participants whose data showed minimal variation, suggestive of stereotyped responses, were excluded, and the rate of diary completion over the 12-month period was calculated.
Fig. 1. (a) Simulated time-series for one participant. An increase in the severity of symptom A appears to precede that of symptom B, and the temporal relationship is suggestive of A driving B. DTVEM looks for the same pattern across the entire time-series, for all participants. (b) Lagged regression between the two symptom time-series produces two coefficients, A to B and B to A. Here, symptom A maximally predicts symptom B over a threshold of statistical significance, with a lag of 3– 17 days. However, symptom B does not significantly predict symptom A.
The predictive lag structure between longitudinal variables was then estimated using DTVEM (Jacobson et al.,2019), which esti-mates the lags at which one variable predicts another, over a threshold of statistical significance from a multi-stage modelling
procedure (Fig. 1). In the exploratory stage, DTVEM uses
generalised additive mixed models [GAMMs, as part of R package mgcv (Wood,2011)] to identify lags that have substantial effects on the outcome variable, thereby narrowing down the search space and improving the computational efficiency of the con-firmatory step. Next, DTVEM passes these to a vector autoregres-sive of order p [VAR( p)] model [using the OpenMx (Neale et al., 2016) R package], which uses a state-space modelling framework to confirm when one variable maximally predicts another. DTVEM cycles iteratively between stages until an optimal solution is found. We only report lag relationships between variables which DTVEM has identified as statistically significant, and also report the combined total effect size of all significant lags between two variables, rather than the effect sizes of each individ-ual lag. DTVEM has been shown to be robust to missing data. This two-stage approach has been shown to be conservative in dramatically reducing type I error, while retaining high power.
In the bivariate, unmediated analyses, a single predictor (sleep quality or duration) and outcome (psychopathology variable) were modelled, to establish the total unadjusted lag relationship (i.e. the total effect between the two variables). Next, the variables were reversed, to examine the reciprocal lag relationship between psycho-pathology and sleep. An auto-regression model was also fitted to examine whether sleep variables predicted themselves over time.
The mediation analyses followed the approach of Baron and Kenny (1986) and used the joint significance test, which has been shown to perform better than many of the alternative meth-ods (Leth-Steensen & Gallitto,2016). First, the putative mediator variable (negative affect or cognitive symptoms) was fitted to the predictor (sleep variable): the absence of a predictive relationship would suggest that the intermediate variable could not act as a mediator. Next, the mediator and predictor were entered together in the same model, with psychosis symptoms as the outcome vari-able. This allows estimation of the direct effect between predictor and outcome in the presence of the mediator, and the indirect effect between predictor and outcome via the mediator. A reduction in the size of the direct effect in comparison with the total effect would support the presence of the mediator, and favour the
Fig. 2.Screenshots of Sleepsight user-facing app: (a) splash page, (b) sleep diary, (c) mood symptoms, (d ) psychosis symptoms. Note that only a section of screens (b–d) are visible.
Fig. 3.Example responses over 12 months from one participant, fitted with a loess smoother with degree 1 polynomial. Higher scores indicate better sleep quality and positive affect, longer sleep duration, and worse negative affect and psychosis symptoms.
mediated model. Also, model fit for the unmediated and mediated analyses was compared using the Bayesian Information Criteria (BIC), which penalises for the complexity of the model and sample size. A lower BIC indicates a closer fit of model to the data. Finally, as negative affect and cognitive symptoms are hypothesised to have a reciprocal relationship on sleep, an oppositional model was exam-ined, where the above steps were repeated with the sleep variable as the mediator and negative affect/cognitive symptoms as the pre-dictor. The effect sizes and model fit of both models were com-pared, and the model which provided the most parsimonious explanation for the data was favoured.
For all analyses, a lag window of 1–20 days was chosen. Given that each analysis is computationally intensive, this strikes a bal-ance between reducing analysis time, while allowing a realistic window of time within which early signs of relapse have been
reported in previous studies (e.g. Birchwood et al., 1989;
Heinrichs & Carpenter,1985; Herz & Melville,1980; Jorgensen, 1998). All variables were standardised to obtain person-centred z-scores, and the number of k selection points was set at 19 (i.e.
lag window−1). Analyses were undertaken in R (R Core Team,
2020), using the Rosalind High Performance Computing
Infrastructure, King’s College London.
Results
Demographic and clinical characteristics
Characteristics of the 36 individuals who took part in the study are presented inTable 1. Six participants were within the first 5 years of their first episode of psychosis and therefore under the care of early intervention mental health services, while the
remaining participants had a more established illness.
Thirty-five out of 36 participants were prescribed regular anti-psychotic medication, of which nine (25%) were taking clozapine for treatment-resistant psychosis, nine (25%) were prescribed more than one antipsychotic and seven (19%) were receiving depot medication. The ISI score for 11 (31%) participants was within the range of clinically significant insomnia (score of 15 or above), and the PSQI score indicated poor sleep quality (score of 5 or greater) for 26 (72%) participants.
The mean (S.D.) duration of participation was 323 (88) days, number of submitted diaries per participant was 246 (136), and the overall response rate was 69%. Data from two participants who were in the study for <1 month, and one participant whose responses showed no variability, were excluded; therefore, data from 33 participants were entered into the final analysis.
Unmediated associations between sleep and psychopathology
Sleep quality negatively predicted psychosis symptoms over lags of 1–8 days (i.e. better sleep quality was associated with lower psych-osis symptoms over an 8-day horizon, and vice versa). It also nega-tively predicted negative affect over lags of 1–8 and 10–11 days, and cognitive symptoms over lags of 1–8 days. Conversely, sleep quality positively predicted positive affect over 1–5 days (i.e. better sleep quality was associated with greater positive affect, over a 5-day win-dow). Effect sizes were in the medium range.
Sleep duration negatively predicted psychosis symptoms over lags of 1–12 days, negative affect over lags of 1–8 days, and cog-nitive symptoms over lags of 1, 3–12 and 14–15 days. No associ-ation between sleep durassoci-ation and positive affect was found (Table 2, top).
Reciprocal associations between psychopathology and sleep variables
Examining the opposite causal direction, statistically significant negative associations of smaller effect size were found between psychosis symptoms and sleep quality (over lags of 1–3 days), negative affect and sleep quality (over lags of 1–3 and 5–6 days), and cognitive symptoms and sleep quality (with lags of 1–2 and 4–5 days). Positive affect predicted better sleep quality over lags of 1–3 and 6–7 days.
No associations were found between psychosis symptoms and sleep duration, positive or negative affect and sleep duration, or cognitive symptoms and sleep duration (Table 2, middle).
Unmediated associations between sleep and individual psychotic symptoms
Sleep quality had a negative relationship with both paranoia and hallucinations, over a lag of 1–5 days, in the low–medium effect size range. In the reciprocal direction, only paranoia, and not hal-lucinations, predicted sleep quality.
Sleep duration predicted hallucinations, but not paranoia, with lags of 2–3 and 5–9 days. No reciprocal association between indi-vidual psychotic symptoms and sleep duration was found.
Auto-regression analyses in sleep and psychopathology variables
Sleep quality predicted itself over lags of 1–5 days (i.e. better sleep quality was associated with better sleep quality over the next five nights), and sleep duration predicted itself over lags of 1–4, 6–8, 10, 14 and 19 days. Negative affect, positive affect, cognitive and psychosis symptoms predicted themselves with lags of 1–4, 1–3, 1–5 and 1–6 days, respectively, with effect sizes in the medium– large range (Table 2, bottom).
Mediated associations between sleep and psychopathology
Results of mediation analyses are set out inTable 3and explained in greater detail in online Supplementary data and Figs S1–S4. In all cases, models where negative affect and cognitive symptoms mediated the relationship between sleep quality/sleep duration and psychosis symptoms were favoured over unmediated relation-ships. Negative affect was a full mediator of the effect between sleep duration and psychosis symptoms, and a partial mediator of the effect between sleep quality and psychosis symptoms. Cognitive symptoms were a partial mediator of the effect between sleep quality and sleep duration, and psychosis symptoms.
Oppositional models where sleep variables mediated the rela-tionship between negative affect/cognitive symptoms and psych-osis symptoms either showed no evidence of an indirect pathway or showed poorer model fit, than the non-mediated total effect, arguing against an indirect relationship mediated by sleep duration or sleep quality.
Discussion Main findings
Sleep disruption has traditionally been considered a consequence of psychosis, arising as the product of mental state disturbance. The possibility that it may play a causal role in the emergence of psychotic symptoms has received support only more recently,
through studies in clinical (Kasanova et al.,2020; Mulligan et al., 2016; Reeve et al., 2018; Waters et al., 2011) and non-clinical populations (Freeman et al., 2017; Hennig, Schlier, & Lincoln, 2019; Reeve, Emsley, Sheaves, & Freeman,2018; Scott, Rowse, & Webb,2017). Using a purpose-built digital platform and novel analytic approach, our findings extend those of recent studies, by determining the temporal relationship over which sleep and psychopathology variables influence and sustain one another.
First, these results support the hypothesis that poorer sleep qual-ity and sleep duration are an antecedent to worsening psychotic symptoms, occurring up to 8 and 12 days prior to the deterioration in mental state, respectively. Poorer sleep quality and duration also anticipate worsening in negative affect and cognitive symptoms, from 1 to 15 days before the deterioration in mood and cognition. Similarly, better sleep quality (but not sleep duration) predicted greater positive affect, but with a shorter lag duration of up to 5 days. Second, we also found evidence of a reciprocal relationship, where greater psychopathology predicted poorer sleep quality (but interestingly, not sleep duration), and greater positive affect predicted better sleep quality. As advanced in previous reports (Harvey, Murray, Chandler, & Soehner,2011; Talbot et al.,2012), this sup-ports the concept of a bidirectional, mutually reinforcing association between sleep quality and psychopathology. Importantly, however, with the exception of positive affect, where the reciprocal association was of longer duration, the lags of reciprocal associations for all other variables were only between 1 and 6 days and of smaller effect size, suggesting that sleep has a more durable and pronounced effect on psychopathology than the opposite direction of the association. This finding is compatible with other studies, which found unidirectional (Kasanova et al.,2020) or only partially reciprocal (Hennig et al., 2019; Reeve et al., 2018) relationships between psychopathology and sleep variables.
It is notable that bidirectional associations were found for sleep quality, but not duration. This provides support to the argument
that changes in the quantity and continuity of sleep drive the devel-opment of mental symptoms, but are in themselves relatively unaffected by mental state fluctuations. One factor that may con-tribute to this finding is the use of sedative antipsychotic medica-tions in all but one of the participants, which have been shown to increase sleep duration and reduce sleep latency (Cohrs,2008; Meyer et al.,2020), thereby dampening the impact of changes in psychopathology on sleep duration. Another explanation for this finding is that the subjective evaluation of sleep duration is less affected by past mental state, whereas sleep quality is more contin-gent on an individual’s sense of well-being. That is, sleep quality is more closely‘tied’ to psychopathology than is sleep duration.
Third, the mediation analyses provided stronger support for models where negative affect and cognitive symptoms mediated the relationship between sleep quality/sleep duration and psych-osis symptoms, than unmediated and oppositional models, where sleep variables were the mediator. This replicates previous findings (Kasanova et al., 2020; Reeve et al., 2018; Scott et al., 2017), and reinforces the view that affective and cognitive pro-cesses lie on the pathway to psychotic symptoms. In exploring the relationship with individual psychotic symptoms, a bidirec-tional relationship between sleep quality and paranoia, and uni-directional association between sleep quality and hallucinations, was found. Also, changes in sleep duration influenced hallucina-tions, but not paranoia. This also accords with previous findings, and supports the hypothesis that the pathway between sleep turbances and paranoia may differ from those linking sleep dis-turbances and hallucinations (Hennig et al., 2019; Reeve et al., 2018; Scott et al.,2017).
Clinical implications and mechanisms
As conceived in the criteria of Bradford Hill (Bradford, 1965), temporality and strength of association are key tenets for inferring Table 1.Demographic and clinical characteristics on study entry (n = 36), and summary statistics of self-reported sleep variables (n = 33)
Mean S.D. Min Max Range
Sex
Male n (%) 24 (67)
Female n (%) 12 (33)
Age/years 40.7 9.7 20.0 59.0 39.0
Duration of illness/years 14.4 8.4 0.8 30.0 29.2
PANSS total score 73.1 16.6 44.0 115.0 71.0
Positive subscale 16.8 4.8 7.0 25.0 18.0
Negative subscale 19.1 7.1 7.0 35.0 28.0
General subscale 37.2 8.5 21.0 57.0 36.0
BDI-II 18.3 11.4 1.0 49.0 48.0
PSQI total score 7.5 4.0 1.0 15.0 14.0
ISI 12.7 6.9 0.0 25.0 25.0
ESS 6.0 5.0 0.0 23.0 23.0
STOP-BANG 2.4 1.6 0.0 6.0 6.0
Self-reported sleep variables
Sleep duration/min 489.8 160.7 0 1425 1425
a causal relationship between a predictor and outcome. In this study, the longer lag duration and larger effect size when sleep variables are predictors, rather than outcome variables, suggest that sleep disturbance emerges earlier in the pathway to deterior-ation of psychosis, and supports a causal role for sleep disturbance in the development of psychopathology. A corollary of this is that subjective sleep disturbances may serve as a clinically useful early warning sign of impending deterioration, and allow the imple-mentation of timely, preventative interventions. Our findings sug-gest that a reduction in sleep duration precedes psychotic deterioration by up to 12 days, suggesting that this is a particularly
important symptom in this regard. A further implication is that psychological and pharmacological interventions which specific-ally target sleep dysfunction may diminish psychopathology and risk of relapse– a strategy which has gained recent experimental support (Freeman et al.,2017).
Another essential criterion for inferring causality is the pres-ence of plausible mechanisms that link predictor and outcome variables. Experimental sleep deprivation in healthy volunteers induces depressed mood and anxiety, executive function and working memory impairment, perceptual distortions and halluci-nations, and paranoia, with particularly strong associations with Table 2.Self-reported sleep variables: lag relationships between single predictor–outcome pairs
Predictor variable Outcome variable Polarity and duration of lag/days Effect sizea
Sleep variables predicting psychopathology
Sleep quality Psychosis symptoms −(1–8) −0.567
Sleep quality Paranoia −(1:5) −0.502
Sleep quality Hallucinations −(1:5) −0.329
Sleep quality Negative affect −(1–8, 10–11) −0.590
Sleep quality Positive affect +(1–5) 0.509
Sleep quality Cognitive symptoms −(1–8) −0.464
Sleep duration Psychosis symptoms −(1–12) −0.501
Sleep duration Paranoia N.S. –
Sleep duration Hallucinations −(2–3, 5–9) −0.332
Sleep duration Negative affect −(1–8) −0.446
Sleep duration Positive affect N.S. –
Sleep duration Cognitive symptoms −(1, 3–12, 14–15) −0.511
Psychopathology predicting sleep variables
Psychosis symptoms Sleep quality −(1–3) −0.203
Paranoia Sleep quality −(1–2, 3, 6) −0.221
Hallucinations Sleep quality N.S. –
Negative affect Sleep quality −(1–3, 5–6) −0.301
Positive affect Sleep quality +(1–3, 6–7) 0.359
Cognitive symptoms Sleep quality −(1–2, 4–5) −0.220
Psychosis symptoms Sleep duration N.S. –
Paranoia Sleep duration N.S. –
Hallucinations Sleep duration N.S. –
Negative affect Sleep duration N.S. –
Positive affect Sleep duration N.S. –
Cognitive symptoms Sleep duration N.S. –
Auto-correlation of sleep and psychopathology variables
Sleep quality Sleep quality +(1–5) 0.625
Sleep duration Sleep duration +(1–4, 6–8, 10, 14, 19) 0.609
Psychosis symptoms Psychosis symptoms +(1–6) 0.865
Negative affect Negative affect +(1–4) 0.783
Positive affect Positive affect +(1–3) 0.707
Cognitive symptoms Cognitive symptoms +(1–5) 0.794
N.S., no significant lag identified.
Table 3.Mediated lag relationships and effect sizes
Model Predictor Mediator Outcome
Total effect, lag
Direct effect, lag
Indirect effect, (path A), lag Indirect effect, (path B), lag BIC, unmediated model BIC, mediated model Preferred model 1a Sleep duration Negative affect Psychosis symptoms −0.501 −(1–12) N.S. (–) −0.446 −(1–8) 0.706 +(1–3) 190 964 61 874 Fully mediated 1b Negative affect Sleep duration Psychosis symptoms 0.704 +(1–3) 0.706 +(1–3) N.S. (–) N.S. (–) 40 454 61 874 Unmediated 2a Sleep quality Negative affect Psychosis symptoms −0.567 −(1–8) −0.072 −(2–4) −0.590 −(1–8, 10–11) 0.688 +(1–3) 190 531 61.691 Partially mediated 2b Negative affect Sleep quality Psychosis symptoms 0.704 +(1–3) 0.688 +(1–3) −0.301 −(1–3, 5–6) −0.072 −(2–4) 40 454 61 691 Unmediated 3a Sleep duration Cognitive symptoms Psychosis symptoms −0.501 −(1–12) −0.054 −(3–6) −0.511 −(1, 3–12, 14–15) 0.613 +(1–2) 190 964 62 147 Partially mediated 3b Cognitive symptoms Sleep duration Psychosis symptoms 0.875 +(1–7) 0.613 +(1–2) N.S. (–) −0.054 −(3–6) 40 065 62 147 Unmediated 4a Sleep quality Cognitive symptoms Psychosis symptoms −0.567 −(1–8) −0.115 −(1–3) −0.464 −(1–8) 0.896 +(1–10) 190 531 61 233 Partially mediated 4b Cognitive symptoms Sleep quality Psychosis symptoms 0.875 +(1–7) 0.896 +(1–10) −0.220 −(1–3, 5–6) −0.115 −(1–3) 40 065 61 233 Unmediated
N.S., no significant lag identified.
Total, direct and indirect effects are calculated from the sum ofβ coefficients over all lags. Refer to online Supplementary Figs S1–S4 for path diagrams.
8 Nicholas Me ye r et al. . https://doi.org/10.1017/S0033291720004857 https://www.cambridge.org/core . IP address: 86.29.239.49 , on 12 Jan 2021 at 09:55:30
mood and cognitive variables (Kahn-Greene, Killgore, Kamimori, Balkin, & Killgore,2007; Petrovsky et al.,2014; Reeve et al.,2018; Waters, Chiu, Atkinson, & Blom, 2018). These effects may be mediated by multiple biological mechanisms, including increased dopamine release and sensitivity (Yates, 2016), impaired sleep-dependent plasticity (Pocivavsek & Rowland,2018) and disrupted synaptic homeostasis (Cirelli & Tononi, 2015). These findings also align with theoretical frameworks which implicate negative affect, worry and rumination (Hartley, Haddock, Vasconcelos e Sa, Emsley, & Barrowclough, 2014), and cognitive impairment (Garety et al.,2013) in driving psychotic symptoms– constructs which were also identified as mediators in this study. There is also growing interest in the role of immune dysregulation in the pathoaetiology of psychosis (Pillinger et al., 2018), which con-verges with evidence suggesting that impaired sleep plays a causal role in the development of metabolic and cardiovascular disease
(Clark et al., 2016), mediated by inflammatory mechanisms
(Irwin, Olmstead, & Carroll, 2016; Mullington, Simpson, Meier-Ewert, & Haack,2010).
Strengths and limitations
A key feature of this study is its use of novel approaches for sampling and analysing intensive longitudinal data that goes beyond single-day lags, and allows the evaluation of longer lag structures between dynamic variables over ecologically valid timescales. There were high levels of engagement with the technology for such a study, yielding a high response rate and an unparalleled dataset.
However, it is important to note that DTVEM operates by exam-ining relationships between multiple concurrent processes in relative time, rather than absolute time, and assumes that the relationship between covariates does not change as a function of study time. Related to this, deterioration in sleep and mental state may occur contemporaneously; however, it is possible the current study design lacks the sensitivity to detect more subtle, early shifts in psychopath-ology. Also, DTVEM remains a novel method, which requires fur-ther validation and replication. The study was promoted as one which examined sleep, meaning that individuals with sleep disor-ders may have been over-represented in our sample. However, rates of insomnia in this study were comparable to those published elsewhere (Palmese et al., 2011; Xiang et al., 2009). We cannot exclude the possibility that the response rate to the daily diary varies with symptom severity at baseline, and also with longitudinal changes in mental state, leading to non-response biases. Finally, given the limitations inherent to self-report measures, including noise and non-response, further examination of the nature of sleep disturbance using objective measures and a better understand-ing of how these measures relate to sleep quality (Della Monica, Johnsen, Atzori, Groeger, & Dijk,2018) are warranted.
Conclusions
Deterioration in sleep quality and sleep duration predicts worsen-ing of psychosis symptoms from up to 8 and 12 days in advance, respectively, and is mediated by negative affect and cognitive symptoms. Psychopathology variables have a reciprocal predictive effect on sleep quality, supporting a bidirectional relationship, but is generally weaker and of shorter duration.
These findings support the conclusion that sleep disturbance plays a causal role in symptom exacerbation and relapse in psych-osis, and suggests it warrants greater attention as an early warning sign of deterioration.
Supplementary material. The supplementary material for this article can be found athttps://doi.org/10.1017/S0033291720004857
Acknowledgements. The valuable contributions of the participants and their clinical teams are gratefully acknowledged. The authors also thank the Clintouch study investigators (University of Manchester) for sharing their vali-dated questionnaire items. The authors are grateful to Dr Ian Grant, Department of Biostatistics and Health Informatics, King’s College London, for support with the High Performance Computing Cluster. The authors acknowledge the use of the research computing facility at King’s College London, Rosalind (https://rosalind.kcl.ac.uk), which is delivered in partnership with the National Institute for Health Research (NIHR) Biomedical Research Centres at South London & Maudsley and Guy’s & St. Thomas’ NHS Foundation Trusts, and part-funded by capital equipment grants from the Maudsley Charity (award 980) and Guy’s & St. Thomas’ Charity (TR130505).
Author contributions. NM designed and conducted the study, analysed data and wrote the manuscript; DWJ and NCJ contributed to the analysis and pres-entation of data; CK developed the data collection platform; DJD, MdV and JHM provided oversight in study design and conduct. All authors contributed to drafting and editing the manuscript.
Financial support. NM is supported by the UK Medical Research Council (grant no. MR/P 001378/1). DWJ is funded by the NIHR Oxford Health Biomedical Research Centre (grant no. BRC-1215-20005). DJD is supported by the UK Dementia Research Institute. This study represents an independent research part funded by the National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care.
Conflict of interest. None.
References
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable dis-tinction in social psychological research. Conceptual, strategic, and statis-tical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. doi: 10.1037/0022-3514.51.6.1173
Bastien, C. H., Vallières, A., & Morin, C. M. (2001). Validation of the insomnia severity index as an outcome measure for insomnia research. Sleep Medicine, 2, 297–307. doi: 10.1016/S1389-9457(00)00065-4
Biagianti, B., Hidalgo-Mazzei, D., & Meyer, N. (2017). Developing digital interventions for people living with serious mental illness: Perspectives from three mHealth studies. Evidence-Based Mental Health, 20, 98–101. doi: 10.1136/eb-2017-102765
Birchwood, M., Smith, J., Macmillan, F., Hogg, B., Prasad, R., Harvey, C., & Bering, S. (1989). Predicting relapse in schizophrenia the development and implementation of an early signs monitoring system using patients and families as observers, a preliminary investigation*. Psychological Medicine, 19, 649–656. doi: 10.1017/S0033291700024247
Bradford, A. (1965). The environment and disease: Association or causation? Proceedings of the Royal Society of Medicine, 58(5), 295–300.
Buysse, D. J., Reynolds, C. F., Monk, T. H., Berman, S. R., & Kupfer, D. J. (1989). The Pittsburgh sleep quality index: A new instrument for psychi-atric practice and research. Psychiatry Research, 28, 193–213. doi: 10.1016/0165-1781(89)90047-4
Cirelli, C., & Tononi, G. (2015). Sleep and synaptic homeostasis. Sleep, 38, 161–162. doi: 10.5665/sleep.4348
Clark, A. J., Salo, P., Lange, T., Jennum, P., Virtanen, M., Pentti, J.,… Vahtera, J. (2016). Onset of impaired sleep and cardiovascular disease risk factors: A longitudinal study. Sleep, 39, 1709–1718. doi: 10.5665/sleep.6098 Cohrs, S. (2008). Sleep disturbances in patients with schizophrenia: Impact
and effect of antipsychotics. CNS Drugs, 22, 939–962. doi: 10.2165/ 00023210-200822110-00004
Della Monica, C., Johnsen, S., Atzori, G., Groeger, J. A., & Dijk, D. J. (2018). Rapid eye movement sleep, sleep continuity and slow wave sleep as
predictors of cognition, mood, and subjective sleep quality in healthy men and women, aged 20-84 years. Frontiers in Psychiatry, 9, 1–17. doi: 10.3389/ fpsyt.2018.00255
Freeman, D., Sheaves, B., Goodwin, G. M., Yu, L. M., Nickless, A., Harrison, P. J.,… Espie, C. A. (2017). The effects of improving sleep on mental health (OASIS): A randomised controlled trial with mediation analysis. The Lancet Psychiatry, 4, 749–758. doi: 10.1016/S2215-0366(17)30328-0
Garety, P., Joyce, E., Jolley, S., Emsley, R., Waller, H., Kuipers, E.,… Freeman, D. (2013). Neuropsychological functioning and jumping to conclusions in delu-sions. Schizophrenia Research, 150, 570–574. doi: 10.1016/j.schres.2013.08.035 Hartley, S., Haddock, G., Vasconcelos e Sa, D., Emsley, R., & Barrowclough, C. (2014). An experience sampling study of worry and rumination in psychosis. Psychological Medicine, 44, 1605–1614. doi: 10.1017/S0033291713002080 Harvey, A. G., Murray, G., Chandler, R. A., & Soehner, A. (2011). Sleep
dis-turbance as transdiagnostic: Consideration of neurobiological mechanisms. Clinical Psychology Review, 31, 225–235. doi: 10.1016/j.cpr.2010.04.003 Heinrichs, D. W., & Carpenter, W. T. (1985). Prospective study of prodromal
symptoms in schizophrenic relapse. American Journal of Psychiatry, 142, 371–373.
Hennig, T., Schlier, B., & Lincoln, T. M. (2019). Sleep and psychotic symp-toms: An actigraphy and diary study with young adults with low and ele-vated psychosis proneness. Schizophrenia Research, 221, 12–19. doi: 10.1016/j.schres.2019.09.012.
Herz, M. I., & Melville, C. (1980). Relapse in schizophrenia. American Journal of Psychiatry, 137, 801–805. doi: 10.1176/ajp.137.7.801
Irwin, M. R., Olmstead, R., & Carroll, J. E. (2016). Sleep disturbance, sleep dur-ation, and inflammation: A systematic review and meta-analysis of cohort studies and experimental sleep deprivation. Biological Psychiatry, 80, 40– 52. doi: 10.1016/j.biopsych.2015.05.014
Jacobson, N. C., Chow, S. M., & Newman, M. G. (2019). The Differential Time-Varying Effect Model (DTVEM): A tool for diagnosing and modeling time lags in intensive longitudinal data. Behavior Research Methods, 51, 295–315. doi: 10.3758/s13428-018-1101-0
Jorgensen, P. (1998). Early signs of psychotic relapse in schizophrenia. British Journal of Psychiatry, 172, 327–330.
Kahn-Greene, E. T., Killgore, D. B., Kamimori, G. H., Balkin, T. J., & Killgore, W. D. (2007). The effects of sleep deprivation on symptoms of psychopath-ology in healthy adults. Sleep Medicine, 8, 215–221. doi: 10.1016/ j.sleep.2006.08.007
Karr, C. (2017). URL: Audacious technologies code for sleepsight app. Retrieved fromhttps://github.com/audaciouscode/sleepsight.
Kasanova, Z., Hajdúk, M., Thewissen, V., & Myin-Germeys, I. (2020). Temporal associations between sleep quality and paranoia across the para-noia continuum: An experience sampling study. Journal of Abnormal Psychology, 129, 122–130. doi: 10.1037/abn0000453
Leth-Steensen, C., & Gallitto, E. (2016). Testing mediation in structural equation modeling: The effectiveness of the test of joint significance. Educational and Psychological Measurement, 76, 339–351. doi: 10.1177/ 0013164415593777
Meyer, N., Faulkner, S. M., McCutcheon, R. A., Pillinger, T., Dijk, D.-J., & MacCabe, J. H. (2020). Sleep and circadian rhythm disturbance in remitted schizophrenia and bipolar disorder: A systematic review and meta-analysis. Schizophrenia Bulletin, 46(5), 1126–1143. doi: 10.1093/schbul/sbaa024. Meyer, N., Kerz, M., Folarin, A., Joyce, D. W., Jackson, R., Karr, C., …
Maccabe, J. (2018). Capturing rest-activity profiles in schizophrenia using wearable and mobile technologies: Development, implementation, feasibil-ity, and acceptability of a remote monitoring platform. JMIR mHealth and uHealth, 6(10), e188. doi: 10.2196/mhealth.8292.
Mulligan, L. D., Haddock, G., Emsley, R., Neil, S. T., & Kyle, S. D. (2016). High resolution examination of the role of sleep disturbance in predicting func-tioning and psychotic symptoms in schizophrenia: A novel experience sam-pling study. Journal of Abnormal Psychology, 125, 788–797. doi: 10.1037/ abn0000180
Mullington, J. M., Simpson, N. S., Meier-Ewert, H. K., & Haack, M. (2010). Sleep loss and inflammation. Best Practice and Research: Clinical Endocrinology and Metabolism, 24, 775–784. doi: 10.1016/ j.beem.2010.08.014
Myin-Germeys, I., Van Os, J., Schwartz, J. E., Stone, A. A., & Delespaul, P. A. (2001). Emotional reactivity to daily life stress in psychosis. Archives of General Psychiatry, 58, 1137–1144. doi: 10.1001/archpsyc.58.12.1137 Neale, M. C., Hunter, M. D., Pritikin, J. N., Zahery, M., Brick, T. R.,
Kirkpatrick, R. M.,… Boker, S. M. (2016). Openmx 2.0: Extended structural equation and statistical modeling. Psychometrika, 81, 535–549. doi: 10.1007/ s11336-014-9435-8
Palmese, L. B., DeGeorge, P. C., Ratliff, J. C., Srihari, V. H., Wexler, B. E., Krystal, A. D., & Tek, C. (2011). Insomnia is frequent in schizophrenia and associated with night eating and obesity. Schizophrenia Research, 133, 238–243. doi: 10.1016/j.schres.2011.07.030
Palmier-Claus, J. E., Ainsworth, J., Machin, M., Barrowclough, C., Dunn, G., Barkus, E.,… Lewis, S. W. (2012). The feasibility and validity of ambulatory self-report of psychotic symptoms using a smartphone software application. BMC Psychiatry, 12, 10. doi: 10.1186/1471-244X-12-172
Palmier-Claus, J. E., Myin-Germeys, I., Barkus, E., Bentley, L., Udachina, A., Delespaul, P. A.,… Dunn, G. (2011). Experience sampling research in indi-viduals with mental illness: Reflections and guidance. Acta Psychiatrica Scandinavica, 123, 12–20. doi: 10.1111/j.1600-0447.2010.01596.x Peeters, F., Berkhof, J., Delespaul, P., Rottenberg, J., & Nicolson, N. A. (2006).
Diurnal mood variation in major depressive disorder. Emotion, 6, 383–391. doi: 10.1037/1528-3542.6.3.383
Petrovsky, N., Ettinger, U., Hill, A., Frenzel, L., Meyhofer, I., Wagner, M.,… Kumari, V. (2014). Sleep deprivation disrupts prepulse inhibition and induces psychosis-like symptoms in healthy humans. Journal of Neuroscience, 34, 9134–9140. doi: 10.1523/JNEUROSCI.0904-14.2014 Pillinger, T., Osimo, E. F., Brugger, S., Mondelli, V., McCutcheon, R. A., & Howes,
O. D. (2018). A meta-analysis of immune parameters, variability, and assess-ment of modal distribution in psychosis and test of the immune subgroup hypothesis. Schizophrenia Bulletin, 45, 1–14. doi: 10.1093/schbul/sby160 Pocivavsek, A., & Rowland, L. M. (2018). Basic neuroscience illuminates causal
relationship between sleep and memory: Translating to schizophrenia. Schizophrenia Bulletin, 44, 7–14. doi: 10.1093/schbul/sbx151
R Core Team (2020). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing.
Reeve, S., Emsley, R., Sheaves, B., & Freeman, D. (2018). Disrupting sleep: The effects of sleep loss on psychotic experiences tested in an experimental study with mediation analysis. Schizophrenia Bulletin, 44(3), 662–671. doi: 10.1093/schbul/sbx103.
Reeve, S., Nickless, A., Sheaves, B., & Freeman, D. (2018). Insomnia, negative affect, and psychotic experiences: Modelling pathways over time in a clinical observational study. Psychiatry Research, 269, 673–680. doi: 10.1016/ j.psychres.2018.08.090
Scott, A. J., Rowse, G., & Webb, T. L. (2017). A structural equation model of the relationship between insomnia, negative affect, and paranoid thinking. PLoS ONE, 12, 1–23. doi: 10.1371/journal.pone.0186233
Talbot, L. S., Stone, S., Gruber, J., Hairston, I. S., Eidelman, P., & Harvey, A. G. (2012). A test of the bidirectional association between sleep and mood in bipolar disorder and insomnia. Journal of Abnormal Psychology, 121, 39– 50. doi: 10.1037/a0024946
Waters, F., Chiu, V., Atkinson, A., & Blom, J. D. (2018). Severe sleep depriv-ation causes hallucindepriv-ations and a gradual progression toward psychosis with increasing time awake. Frontiers in Psychiatry, 9, 303. doi: 10.3389/ fpsyt.2018.00303.
Waters, F., Sinclair, C., Rock, D., Jablensky, A., Foster, R. G., & Wulff, K. (2011). Daily variations in sleep-wake patterns and severity of psychopathology: A pilot study in community-dwelling individuals with chronic schizophrenia. Psychiatry Research, 187, 304–306. doi: 10.1016/j.psychres.2011.01.006 Wood, S. N. (2011). Fast stable restricted maximum likelihood and marginal
likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society (B), 73, 3–36.
Xiang, Y. T., Weng, Y. Z., Leung, C. M., Tang, W. K., Lai, K. Y., & Ungvari, G. S. (2009). Prevalence and correlates of insomnia and its impact on quality of life in Chinese schizophrenia patients. Sleep, 32, 105–109.
Yates, N. J. (2016). Schizophrenia: The role of sleep and circadian rhythms in regulating dopamine and psychosis. Reviews in the Neurosciences, 27, 669–687.https://dx.doi.org/10.1515/revneuro-2016-0030
